I've been waiting for Stephen to drop his big idea, since I'm always up for considering the next big thing. I understand this as follows (and Stephen correct me if I misinterpret you): companies don't need just applications. They need applications and human resources to integrate and deploy applications. Providers like IBM offer one stop shopping for applications and services, but why not allow the resources to collaborate in a more ad hoc fashion with a thin mediator (Stephen's network app) in the middle? Get IBM out of the picture all together.
Another aspect to Stephen's proposal is packaged applications which are ready to deploy. After users select their application, they select developers to go along with it. This got me thinking. OpenSource projects do a lousy job promoting thirdparty integrators. The classic question OS developers ask is "how do I make money on my OpenSource project?" Stephen might offer a solution. Allow developers and implementors to sign up as consultants and advertise them off the home page of your project. Take a fee when companies buy implementation resources. The project becomes a conduit between users with money and consultants.
Nagios is a classic example of this. It is more of a platform than a final product. It requires implementors to set up and test. I've done it, but I'm pretty sure we would have been better off hiring a specialist.
This seems like something SpikeSource or SourceForge could be doing right now.